This study focuses on revealing and developing personal constructs regarding problem behavior in classrooms. 54 teachers (initial and in-service trainees) took part in this project. The main idea is that teachers' opinions about their students and themselves influence the way they interact with them. Their thoughts and ideas about students - their personal constructs - are generally unconscious. These are formed by one's life history (experiences in childhood, at school, sporting clubs etc.). By making them explicit, teachers can find out which problem-behavior-type they especially like or dislike. To reveal teachers' constructs in their perceptions of their students, we used the repertory grid technique, invented by Kelly (1955). Kelly argued that each individual uses a unique set of personal constructs in interpreting and predicting events. Those constructs tend to take an either-or form (for example involved-detached; independent - helpless). Two teachers may use different constructs in evaluating the same student and these differences lead to a different approach. Therefore it is important to know and to learn about one's own constructs and those of colleagues. To develop and positively influence personal constructs on problem behavior, we chose a developmental reflective approach (Heikkinen, 1998; Pajak, 2000). In this coaching- approach, attention is paid to reflection on recollection and to a theoretical orientation towards problem behavior. Both the repertory grid-approach and the coaching model give a powerful impulse to the development of thinking and acting of teachers, because of the focus on consciousness of personal constructs, the relation between the past en the present, the own strengths and vulnerabilities and theoretical information on behavior problems. It can help teachers to build their identity towards 'problemchildren'In our research we focus on:- the change in personal constructs on students over time: do they become more differentiated (not only social-emotional, but also attitudes towards work, achievement etc.) and more related to known indicators for school success, such as cooperation and self-control skills (Lane, Pierson & Givner, 2003).- the training-conditions that affect the results (Kelly's repertory grid technique, coachingsmodel etc.).We will discuss the quantitative and qualitative aspects of the research project in the framework of initial and continuing teacher education (regular and special).